You are on page 1of 5

Psychological Medicine, 1993, 23, 87-91.

Copyright © 1993 Cambridge University Press

Working memory in clinical depression: an


experimental study
SHELLEY CHANNON,' JANE E. BAKER AND MARY M. ROBERTSON
From the Department of Academic Psychiatry, University College and Middlesex School of Medicine,
London

SYNOPSIS This study compared clinically depressed subjects with normal controls on a range of
working memory tasks. The findings suggested the articulatory loop and visuospatial sketch pad
components of working memory to be unimpaired in depression. On a range of clinical tasks likely
to involve central executive function, depressed subjects showed impairment only on some tasks.

demands on attentional resources, or where


INTRODUCTION performance can be improved by active encoding
Empirical research with both dysphoric and strategies. In terms of the Baddeley & Hitch
clinically depressed subjects has demonstrated (1974) working memory model, deficient al-
convincingly that depression is associated with location of processing resources associated with
deficits in aspects of long-term memory (for a depression might be expected to reduce the
review see Williams el al. 1988). Clinical ex- capacity of the central executive, but the
perience suggests that depressed patients also relatively passive processing of the articulatory
report problems with attention and concen- loop and visuospatial sketch pad should not be
tration, although this has received less con- affected.
sideration by researchers. Some studies have In the following study, these predictions were
been carried out, but not always in the context investigated by comparing clinically depressed
of a clearcut theoretical framework. subjects with normal controls on a range of
The model of working memory proposed by tasks selected to involve aspects of working
Baddeley & Hitch (1974) provides a theoretical memory. The articulatory loop was studied
framework which allows specific predictions to using tasks taken directly from the working
be made. They postulated a central executive, a memory research, while the other measures were
processor which allocates attentional resources selected because of their widespread use in
and mediates problem-solving and decision- clinical practice.
making, and two subordinate systems, the
articulatory loop and the visuospatial sketch METHOD
pad. The supervisory attentional system postu-
lated by Norman & Shallice (e.g. Shallice, 1988) Subjects
is assumed to be similar in function to the DSM-III-R criteria (American Psychiatric As-
central executive (Baddeley, 1988). sociation, 1987) were used to select right-handed
Theoretical accounts of the effects of de- subjects who reached diagnostic criteria for
pression on cognitive function have commonly Major Depression, and had no history of alcohol
related these to deficient allocation of attentional or drug abuse, physical illness or other psy-
resources (e.g. Ellis & Ashbrook, 1988; Hertel & chiatric disorder. The depressed group com-
Hardin, 1990). Such models contend that de- prised 24 subjects (7M, 17F) of mean age 39-29
pressive deficits occur in tasks which make high (s.D. 11-09), 15 of whom were prescribed anti-
depressant medication at the time of testing.
1
Address for correspondence: Dr S. Channon, Wolfson Building,
Similar exclusion criteria were used to select 21
Middlesex Hospital, Mortimer Street, London WIN 8AA. non-depressed control subjects (6M, 15F), of
87
S. Channon and others

mean age 3900 (S.D. 13-94). The t test showed Word length effect This task was adapted
that the groups did not differ significantly in from that described by Baddeley et al. (1984),
age; nor did they differ significantly in sex ratio designed to investigate the length of the articu-
using the x2 test. latory loop. Subjects listened to a tape recording
of 2 practice lists and 20 trial lists of words, each
Clinical measures list containing 5 short (1 syllable) or long (5
All subjects underwent a clinical interview. syllable) words. After each list, they were asked
Depression was assessed using the Hamilton to write down the words in order of presentation.
Depression Scale (Hamilton, 1967) and the Beck The trials were divided into 2 sets of 10, and
Depression Inventory (Beck et al. 1961). In each set contained 5 short and 5 long trials. For
addition, the State-Trait Anxiety Inventory 1 of the 2 sets of trials, subjects were asked to say
(Spielberger et al. 1970) and the Ley ton Obses- 'one, two' repeatedly while listening to the list
sional Inventory (Snowdon, 1980) were also and writing it down (suppression condition).
administered. The depressed subjects scored Half the subjects in each group performed in the
significantly higher than the controls on each of suppression condition for the first trial set, and
these measures at the P < 0-0001 level, with the half for the second trial set. This gave 4 types of
exception of trait obsessionality, which was trials: short words with and without suppression
significant at the P < 0001 level. and long words with and without suppression.
The Vocabulary and Block Design subtests of
the WAIS-R (Wechsler, 1981) were administered Forward digit span Span of immediate recall
as indices of verbal and non-verbal general of digits using digit span from the WAIS-R
ability. Mean age-scaled scores on the (Wechsler, 1981).
Vocabulary subtest were 12-13 (S.D. 3-02) and
Visuospatial sketch pad
1305 (S.D. 2-46) for the depressed and control
groups respectively, and on the Block Design Forward block sequence span Span of im-
subtest 11-50 (S.D. 2-78) and 11-90 (s.D. 311) mediate recall of spatial sequences using Corsi's
respectively. The groups did not differ signifi- block sequence span, described by Milner (1971).
cantly on these measures.
Tasks involving the central executive
Experimental measures Backward digit span and block sequence span
Articulatory loop Backward recall of sequences using digit and
block sequence span (described above).
Phonological similarity effect This test was
adapted from the task described by Wilding & Paced and unpaced auditory serial addition test
Mohindra (1980), designed to investigate the {PASAT) This task presented a series of 31
phonological store of the articulatory loop. audiotaped digits, and required subjects to add
Each trial consisted of a list of 5 letters presented pairs of digits so that each new digit was added
on a computer screen at a rate of 1 letter per sec. to the immediately preceding one (Gronwall &
There were 2 possible letter sets, either CDGPT Sampson, 1974). After practice, subjects first
(acoustically confusable) or HJMRZ (acous- performed an unpaced task, where they were
tically non-confusable), each in 12 different given the next digit as soon as they had given the
random (non-meaningful) sequences. Subjects answer to the sum of the previous 2 digits.
were instructed to watch each trial presentation Following Gotham et al. (1988), 2 paced series
carefully, and then write down the letters in were then presented, with inter-trial intervals of
order of presentation. In addition, in half the 4 and 2 sec respectively.
trials they were required to say 'one, two'
Trail-making test This test, described by Reitan
repeatedly while the letters were appearing on
(1958), requires subjects initially to join numbers
the screen (suppression condition). There were
in ascending order, and then alternate between
thus 4 test conditions in all, each comprising 6
letters and numbers, as fast as possible.
trials: confusable letters with and without sup-
pression and non-confusable letters with and Letter cancellation test In this task subjects
without suppression. Order of presentation was were required to cancel out visually presented
counter-balanced within groups. target letters which were randomly interspersed
Working memory in depression 89

in rows of distractor letters, based on the similar pattern was found in the present study
vigilance tasks developed by Diller et al. (1974). using repeated measures analysis of variance
Identical stimuli were used for 2 conditions, (ANOVA) with one between-subjects factor
crossing out a single target ('H'), and then 2 (group) and two within-subjects factors (confus-
targets (' C 's a n d ' E 's) as fast as possible without ability, suppression). Thus, there was a sig-
errors. nificant list by suppression interaction across
groups ( F = 22-31, df = 1,42, P < 00001). Sub-
Procedure sequent analyses showed that the difference
After giving informed written consent, subjects between confusable and non-confusable lists did
were given the clinical interview; they then not reach significance in the suppression con-
carried out the experimental tasks and clinical dition, but did in the non-suppression condition
measures, with appropriate breaks between tests. (F= 59 51, df = 1,42, P < 00001). There were
Order of presentation of tests was counter- no significant two- or three-way group inter-
balanced to control for any effects of fatigue. In actions with group, nor a significant main effect
addition to these tests subjects also participated of group.
in a wordlist learning task comparing memory
for materials with different levels of structure, to Word length effect
be reported separately. Previous work reported that in normal subjects,
there was no difference between long and short
Scoring
words in the suppression condition, and a
The data for the phonological similarity effect significant difference between long and short
tasks were scored by adding the number of words in the non-suppression condition (Bad-
correct trials out of a possible 6 for each of the deley et al. 1984). In the present study, the data
4 conditions. Similarly, for the word-length task
the data were scored in terms of the number of
correct trials out of a possible 5 for each of the Table 1. Means for the depressed and control
four conditions. For the digit and block sequence groups on the experimental and self-report
spans, the score consisted of the span length of measures
digits or blocks correctly reproduced forwards
Depressed Controls
or backwards. The PASAT was scored in terms
of the number of errors made out of a possible Experimental measures Mean (s.D.) Mean (s.D.)
30 on each of the 3 tasks. For the trail-making
Phonological similarity effect
test and the letter cancellation tasks, scores were (correct trials)
the speed of performance in each condition. Confusable/suppression 1 % (1 46) 2-14(1-77)
Confusable/non-suppression 2-39(1-92) 3-24(1-51)
Non-confusable/suppression 2-48(1-78) 2-05(1-72)
RESULTS Non-confusable/non-suppression 4-78(1-54) 505(102)
Word length effect (correct trials)
The mean scores and standard deviations for the Short words/suppression 0-77(1-11) 1-38(1-36)
Short words/non-suppression 2-36(1-87) 3-19(1-60)
depressed and control groups for each measure Long words/suppression 0-64(1-18) 0-95(1-50)
are shown in Table 1. A significance level of 5 % Long words/non-suppression 1-14(1-39) 1-52(1-54)
was adopted throughout, with appropriate alpha Forward digit span 6-92(1-35) 7-33(1-20)
adjustment (alpha divided by number of tests) Backward digit span 508(1-25) 6-38(1-20)
Forward block span 5-55(101) 5-71 (101)
when post hoc tests were performed. Backward block span 5-05(1-13) 5-48(1-33)
PASAT errors
Articulatory loop 2-sec presentation 11-95(5-70) 9-25(6-91)
Phonological similarity effect 4-sec presentation 4-05 (5-42) 3-35 (482)
Unpaced presentation 5-19(5-51) 2-85(3-10)
Previous research showed that with normal
Trail-making (s)
subjects, performance on the confusable and Numbers alone 43-20 (29-56) 34-19 (12-40)
non-confusable lists did not differ significantly Numbers and letters 88-50 (7006) 69-48 (33-96)
in the suppression condition, but that there was Letter cancellation (s)
One target 60-00(12-67) 61-71 (17-60)
a significant difference between lists in the non- Two targets 115-64(24-28) 11519(26-49)
suppression condition (e.g. Murray, 1968). A
90 S. Channon and others

were transformed logarithmically to reduce and backward digit span and PASAT per-
skewness. Repeated measures ANOVA revealed formance, the experimental measures which
a significant word length by suppression in- discriminated best between the groups. There
teraction across groups (F = 5-96, df = 1,41, P were no significant correlations with the self-
< 0019). Separate analyses for the suppression report measures for either group. Nor were there
and non-suppression conditions using a strict significant associations within the depressed
significance level (alpha/2 = 0-025) showed a group with past history of depression, past
significant difference between long and short history of dysthymia, scores on the Hamilton
words in the non-suppression condition (F = Scale, or presence or absence of anti-depressant
30-77, df = 1,41, P< 00001), and a near- medication. A t test comparison of the medicated
significant difference in the suppression con- and unmedicated depressed subjects on these
dition CF = 4-65, df = 1,41, />< 0-037). There tasks showed that the groups did not differ
were no significant two- or three-way inter- significantly on either backward digit span
actions with group, nor a significant effect of (t = 008, P = 0-935) or the PASAT measures
group. (unpaced: t = 0-48, P = 0-636; 4-sec; / = 1-74,
P = 0099; 2-sec: t = 0-40, P = 0-696).
Forward digit span
The t test revealed no group difference on this DISCUSSION
measure.
The main finding was that on tasks involving
Visuospatial sketch pad aspects of working memory, clinically depressed
Forward block sequence span and control subjects did not differ on several
Again the / test revealed no group difference on measures involving the articulatory loop, or on
this measure. one measure involving the visuospatial sketch
pad. In a range of clinical tasks involving the
Tasks involving the central executive central executive, the groups differed sig-
Backward digit span and block sequence span nificantly only on backward digit span, and
Analysis by t test showed that the depressed approached a significant difference on the
subjects recalled significantly fewer digits back- PASAT, with poorer performance by the de-
wards than controls (t = 3-54, P < 0001). The pressed subjects. While it might be hypothesized
that antidepressant medication could produce
groups did not differ significantly on backward
block sequence span.
impairments in working memory, comparison
of medicated and unmedicated depressed sub-
PASAT jects in the present study did not support this
explanation.
These data were transformed logarithmically to The experimental findings were consistent
reduce skewness. Repeated measures ANOVA with the prediction based on theoretical models
comparing the groups on the three tasks, of cognitive function in depression (e.g. Ellis &
unpaced, 4-sec and 2-sec serial addition, did not Ashbrook 1988; Hertel & Hardin, 1990) that
show a significant group x task interaction; there depressed subjects should not show impairments
was a near-significant effect of group (F = 3-84, in the articulatory loop or visuospatial sketch
df= 1,39, P < 006). pad. Thus, the depressed group did not differ
significantly from normal controls on tasks
Trail-making test and letter cancellation measuring the phonological storage of infor-
Repeated measures ANOVA revealed no sig- mation (phonological similarities task), or the
nificant group x task interactions or effects of length of the rehearsal loop (word length task).
group on either of these tests. The groups also did not differ significantly on
forward digit span, which also involves the
Relationship to indices of mood articulatory loop, or on forward block sequence
Pearson product moment correlations were span, which involves the visuospatial sketch
calculated within each group to examine any pad.
relationship between the self-report measures It was also predicted that depressed subjects
Working memory in depression 91

might be expected to show impairment on tasks


involving the central executive, and there was REFERENCES
some evidence for this. The depressed subjects American Psychiatric Association (1987). Diagnostic and Statistical
did show impairment on backward digit span, Manual of Mental Disorders, 3rd edn., revised. American Psy-
chiatric Association: Washington, DC.
which requires the manipulation of sequences in Baddeley, A. (1988). Cognitive psychology and human memory.
the central executive in addition to temporary Trends in Neuroscience 11, 176-181.
Baddeley, A. & Hitch, G. J. (1974). Working memory. In Recent
storage. Warren & Groome (1984) reported Advances in Learning and Motivation, vol. VIII (ed. G. Bower), pp.
similar findings in clinically depressed subjects 47-90. Academic Press: New York.
before ECT, with no impairment on forward Baddeley, A., Lewis, V. & Vallar, G. (1984). Exploring the
articulatory loop. Quarterly Journal of Experimental Psychology
digit span and marginal impairment on back- 36A, 233-252.
ward span. The near-significant difference be- Beck, A. T., Ward, C. H., Mendelson, M., Mock, J. & Erbaugh, J.
tween the groups on the PASAT tasks in the (1961). An inventory for measuring depression. Archives of General
Psychiatry 4, 561-571.
current study is also consistent with expecta- Diller, L., Ben-Yishay, Y., Gerstman, L. J., Goodkin, R., Gordon,
tions, since this involves the temporary storage W. & Weinberg, J. (1974). Studies in Cognition and Rehabilitation
of a sequence of numbers in the articulatory in Hemiplegia (Rehabilitation Monograph No. 50). New York
University Medical Center Institute of Rehabilitation Medicine:
loop, and serial addition of consecutive pairs of New York.
these numbers in the central executive. However, Ellis, H. C. & Ashbrook, P. W. (1988). Resource allocation model of
the effects of depressed mood states on memory. In Affect,
the depressed subjects were unimpaired on Cognition and Social Behaviour (ed. K. Fiedler and J. Forgas), pp.
several other measures likely to involve the 25-43. Hogrefe: Toronto.
central executive, namely backward block se- Gotham, A. M., Brown, R. G. & Marsden, C. D. (1988). 'Frontal'
cognitive function in patients with Parkinson's disease ' o n ' and
quence span, the trail making test, and letter 'off' levodopa. Brain 111, 299-321.
cancellation. Gronwall, D. M. A. & Sampson, H. (1974). The Psychological Effects
of Concussion. Auckland University Press: Auckland.
According to the models which postulate Hamilton, M. (1967). Development of a rating scale for primary
deficient attentional resources (e.g. Ellis & depressive illness. British Journal of Social and Clinical Psychology
Ashbrook, 1988; Hertel & Hardin, 1990), de- 6, 278-295.
Hertel, P. T. & Hardin, T. S. (1990). Remembering with and without
pressive deficits should occur maximally on awareness in a depressed mood: evidence of deficits in initiative.
tasks which make high demands on attentional Journal of Experimental Psychology: General 119, 45-59.
resources, or where performance can be im- Milner, B. (1971). Interhemispheric differences in the localisation of
psychological processes in man. British Medical Bulletin 27,
proved by active encoding strategies. It may be 272-277.
that the pattern of findings in the present study Murray, D. J. (1968). Articulation and acousticconfusability. Journal
reflects the differential sensitivity of the various of Experimental Psychology 78, 679-684.
Navon, D. (1987). The role of outcome conflict in dual-task
clinical tasks to any reduction of available interference. Journal of Experimental Psychology: Human Per-
processing resources in the central executive. ception and Performance 13, 435-448.
However, any post-hoc explanations on this Reitan, R. M. (1958). Validity of the Trail Making Test as an
indication of organic brain damage. Perceptual and Motor Skills 8,
basis can only be speculative in nature. Alterna- 271-276.
tively, an explanation might be sought in terms Shallice, T. (1988). From Neuropsychology to Mental Structure.
Cambridge University Press: Cambridge.
of multiple processing resources (e.g. Wickens, Snowdon, J. (1980). A comparison of written and postbox forms of
1984; Navon, 1987), with the various tasks the Leyton Obsessional Inventory. Psychological Medicine 10,
drawing on specialized processing resources to 165-170.
differing extents. Spielberger, C. D., Gorsuch, R. L. & Lushene, R. E. (1970). Manual
for the State-Trait Anxiety Inventory. Consulting Psychologists
Overall, the current findings suggest that any Press: Palo Alto, CA.
Warren, E. W. & Groome, D. H. (1984). Memory test performance
effects of clinical depression on working memory under three different waveforms of ECT for depression. British
are likely to spare the articulatory loop and Journal of Psychiatry 144, 370-375.
visuospatial sketch pad. The evidence for im- Wechsler, D. (1981). Wechsler Adult Intelligence Scale - Revised.
Manual. Psychological Corporation: New York.
pairment of the central executive was patchy, Wickens, C. (1984). Processing resources in attention. In Varieties of
and further studies are needed which system- Attention (ed. R. Parasuraman and D. R. Davies), pp. 63-102.
atically explore the effects of varying the Academic Press: New York.
demands on central executive function. Wilding, J. & Mohindra, N. (1980). Effects of subvocal suppression,
articulating aloud and noise on sequence recall. British Journal of
Psychology 71, 247-261.
Williams, J. M. G., Watts, F. N., MacLeod, C. & Mathews, A.
This study was funded by the North East Thames (1988). Cognitive Psychology and Emotional Disorders. Wiley:
Locally Organized Research Scheme. Chichester.

You might also like